• Text in Japanese

FUJIFILM Corporation (President: Kenji Sukeno, Fujifilm) and Osaka University (President: Shojiro Nishio) are setting up a Joint Research Chair^*1 within the Osaka University’s Graduate School of Medicine in April 2019. Fujifilm and Osaka University will work to develop advanced Artificial Intelligence (AI) technologies, such as AI-based medical diagnostic imaging support system and AI that can explain thought processes behind determining lesion locations. The AI that can explain these thought processes is called Explainable AI or XAI^*2.

[Objectives of setting up the Joint Research Chair]
In recent years, the use of AI technology has been attracting growing attention in the field of medical diagnostic imaging. Expectations are high on research and development (R&D) of a system that incorporates an algorithm capable of detecting and differentiating lesions at a high level of accuracy.

The Artificial Intelligence Diagnostic Imaging Joint Research Chair to be set up this time will work on building a massive database of medical images, captured with high-performance equipment of Osaka University Hospital and combined with annotation information^*3 added by diagnostic imaging specialists. Such a high-quality and large-scale database is essential in R&D of high-precision AI technology that assists medical diagnostic imaging. The database will be used by the two partners to promote R&D and social implementation of cutting-edge AI technology that assists medical diagnostic imaging. The project strives to develop and achieve practical application of more advanced AI technologies such as XAI and X-shot learning^*4, which shows accurate performance in identifying potential diagnosis candidates even from very limited amount of case data.

[Diagnostic imaging research achievements of Osaka University]
Professor Noriyuki Tomiyama of Osaka University Graduate School of Medicine (Diagnostic and Interventional Radiology) is conducting an advanced level of clinical research and education, using precision diagnostic imaging devices such as latest multi-row X-ray detector CT scan^*5 and high-field MRI^*6. In diagnostic imaging, Osaka University has excellent diagnostic imaging specialists in the head, chest and abdominal categories to offer highly specialized medical consultation and high-quality diagnostic imaging reports. The University has made numerous research achievements in computer-aided diagnosis, such as determination of benign / malignant pulmonary nodules and liver segmentation using CT scans. These achievements have contributed to the advancement of AI research.

[Fujifilm R&D in AI diagnostic imaging ]
Fujifilm has applied its organ segmentation technology, developed with machine-learning techniques, mainly to the preoperative simulation function of 3D image analyzers. The company is currently carrying out in-house development of AI technologies that can assist medical diagnostic imaging, support workflow at the medical frontline, and be utilized in the field of medical device maintenance services. It is also partnering with AI technology vendors with outstanding technological expertise in and outside Japan to develop solutions that streamline diagnostic workflow.

Fujifilm will provide these AI technologies globally under the “REiLI” brand as solutions catering to the needs and workflows of various markets.

🔹 Description of terms

*1 Joint research chair
A joint research chair is an independent research organization created within a university, using private-sector funds. It accepts researchers and research resources from the company funding the program, in which university academics and researchers from the company work together on an equal basis to carry out joint research on a common theme. The program is designed to work toward producing excellent research outcomes that have practical applications in society.

*2 XAI
XAI is AI technology that can present reasons of decision-making to humans, so that it can be applied to fields that require transparency and “trust” from users. In medical imaging, XAI can be used, for example, to explain which part or pattern of images have been used to determine lesion locations.

*3 Annotation information
Information added to data as annotation, such as comments on the location of suspected ailment

*4 X-shot learning
Humans have the ability to identify the characteristics of just one or only a small number of cases, remembering it as a specific category and recognizing similar cases. X-shot learning (or one-shot learning) refers to a task that enables learning from single or an extremely small amount of sample data.

*5 Multi-row X-ray detector CT scan
This is a CT scan system that uses a detector featuring X-ray elements in multiple rows to obtain a scan across a wide breadth each time. It provides clear images eliminating effects of body movements and breathing during scanning. It is also capable of 4D CT images that include time axis.

*6 High-field MRI
This is an MRI system that scans in a static magnetic field using superconducting magnets with a higher field strength. It can mitigate the impact of noise on signals, producing more detailed and sharp images.

[Overview of the Joint Research Chair]